Power-law singularity in the local density of states due to the point defect in graphene Wen-Min Huang, 1 Jian-Ming Tang, 2 and Hsiu-Hau Lin 1,3 1 Department of Physics, National Tsing-Hua University, Hsinchu 300, Taiwan 2 Department of Physics, University of New Hampshire, Durham, New Hampshire 03824-3520, USA 3 Physics Division, National Center for Theoretical Sciences, Hsinchu 300, Taiwan Received 25 July 2009; published 17 September 2009 Defects in graphene give rise to zero modes that are often related to the sharp peak in the local density of states near the defect site. Here we solved all zero modes induced by a single defect in the finite-size graphene and show that their contributions to the local density of states vanish in the thermodynamic limit. Instead, lots of resonant states emerge at low energies and eventually lead to a power-law singularity in the local density of states. Our findings show that the impurity problem in graphene should be treated as a collective phenomenon rather than a single impurity state. DOI: 10.1103/PhysRevB.80.121404 PACS numbers: 81.05.Uw, 71.20.-b, 71.23.-k, 71.55.-i Graphene, a single-layer graphite composed of carbon at- oms arranged in two-dimensional honeycomb lattice, is re- cently fabricated in laboratory and attracts intense attentions from both experimental and theoretical aspects. 14 One of the striking features of graphene is its relativistic description, described by a pair of massless Dirac fermions in the low- energy regime. 5,6 The relativistic spectrum gives rise to in- teresting phenomena such as half-integer quantum Hall effect, 7 Klein paradox, 8 edge magnetism, 9,10 and others. 1113 In a two-dimensional Dirac system a peculiar quasilocal- ized state can be induced by a point defect. 14 Such a state can be seen in graphene as a pronounced peak in the local den- sity of states LDOSat zero energy. 1517 Density-functional studies 18,19 suggest that point defects can be created by chemisorption of hydrogen atoms. 20 Furthermore, there exists quantized magnetic moment associated with each defect. 18,21,22 Is the peak in the LDOS near the defect site caused by the zero modes in graphene? It is tempting to say yes. However, we revisit this problem and find the origin of the peak is not from the zero modes. We start from the finite-size graphene in nanotorus geometry and investigate how the system evolves toward the thermodynamic limit. At finite system size N, we can solve the zero modes due to a single defect analytically, contributing a zero-energy peak in LDOS as expected. 23 However, its spectral weight decays as 1 / ln N or 1 / N depending on whether the nanotorus is semiconducting or metallicand eventually vanishes in two dimensions. Therefore, even though the defect gives rise to zero modes, they do not contribute to the pronounced peak in LDOS for two-dimensional graphene. The pronounced peak in LDOS found in previous studies can be explained in two steps. First of all, our numerics show that the defect in graphene induces enormous resonant peaks in the LDOS at energies close to zero. Then, as the system size grows to infinity, these peaks crowd into zero energy and become singular. Both numerical and analytic ap- proaches give 1 / |E| power-law singularity with weak loga- rithmic corrections. That is to say, the peak in the LDOS is not from a single impurity state. Instead, it is a power-law singularity from collective resonance induced by the defect. It is rather amusing that the impurity state in graphene dis- solve into a power-law singularity as the single-particle state disappears in one-dimensional interacting electron gas. 24 The emergence of the power-law singularity resembles the quan- tum criticality found in many two-dimensional systems 25,26 and suggest that graphene is a quantum critical system 27,28 as well. As a result, introducing a point defect reshuffles the LDOS leading to a power-law singularity rather than a delta- function or broadened Lorentzianpeak. Now we walk through the details which lead to the con- clusions sketched in above. Because the band structure of graphene obtained by the first-principles calculations is well approximated by the nearest-neighbor hopping for the active orbitals, 29,30 it is sufficient to start from a tight-binding Hamiltonian and add a single defect at the origin, H =- t r,r' c rcr'+ c r'cr + V 0 c 0c0, 1 where t is the nearest-neighbor hopping amplitude, and V 0 is the strength of the impurity potential. In the remaining part of the calculations, we mainly focus on the unitary limit V 0 . Though the impurity state in the unitary limit has been solved analytically in a recent paper, 15 it is insightful to red- erive it with cares so that the evolution of the impurity state with the system size N is clarified. For simplicity, we apply periodic boundary conditions in both directions, wrapping the finite-size graphene into a nanotorus. The number of unit cells along the x and y axes is N x and N y and the lattice sites are labeled by the coordinates x and y = n + x / 2 with x =0, 1,..., N x / 2 and n =0,1,..., N y -1 as shown in Fig. 1. Let us consider the carbon nanotube limit first by taking N x but keeping N y finite. Because the honeycomb lattice is bipartite, the wave function of the zero modes only show up in one of the sublattices 31 as shown in Fig. 1. The wave function for the left and right sectors can be written as the linear combinations of all evanescent modes, l x, y= l C l e ik l y z l -x , r x, y= r C r e ik r y z r x , 2 PHYSICAL REVIEW B 80, 121404R2009 RAPID COMMUNICATIONS 1098-0121/2009/8012/1214044©2009 The American Physical Society 121404-1